1. Technical Field
The present invention relates generally to measurement systems, and more particularly, to determining a matching problem of a measurement system under test to a fleet including at least one other measurement system, and also determining a root cause issue of the matching problem.
2. Related Art
Measurement systems are applied in a variety of industries in which precise and accurate measurements are required, such as semiconductor manufacturing. Challenges relative to attaining quality measurement are presented in terms of individual measurement systems and across a fleet of measurement systems.
In terms of individual measurement systems, each tool is typically required to achieve small tolerances to achieve better quality products and fewer rejections in the manufacturing process. For example, in the semiconductor manufacturing industry, the 1999 Edition of the International Technology Roadmap for Semiconductors (ITRS precision specification) lists the necessary precision needed for isolated line control in the year 2001 to be 1.8 nm. Correctly assessing and optimizing the measurement potential of a measurement system is difficult for a number of reasons. For example, an evaluator normally has limited access to the various instruments under consideration. In addition, each instrument needs to be evaluated under a wide range of conditions in order to gain a valid impression of how it will perform in the actual manufacturing setting. Finally, there are no widely accepted standards relative to the required parameters and how the parameters should be measured. One approach, disclosed in PCT Publication WO/2004/059247, which is hereby incorporated by reference, involves assessing and optimizing a measurement system by determining a total measurement uncertainty (TMU) based on precision and accuracy. The TMU is calculated based on a linear regression analysis and removing a reference measuring system uncertainty (URMS) from a net residual error. The fundamental question answered in the TMU PCT publication is how to correct or accurately determine how the measurement system under test or fleet under test measures. The TMU publication, however, does not address how similarly the measurement system under test matches the reference measurement system.
When quality measurement is evaluated across a fleet of measurement systems, the above-described challenges for assessing and optimizing a single meteorology tool are multiplied. The ITRS precision specification referred to in the previous paragraph actually applies to whatever set of tools is used to monitor and control critical steps in the semiconductor manufacturing process. It is more cost effective to avoid dedicating tools to specific manufacturing steps by allowing any tool of the full fleet in the manufacturing line to make measurements. This, however, places great demand on achieving and maintaining good measurement matching for all tools in the fleet. Typically, measurement systems having similar measurement technology are selected for use together. Then, the measurement systems across a fleet are preferably manually matched as much as possible. In order to achieve matching, in one approach, an average offset value between measurements of tools within a fleet is minimized to match the tools as much as possible. A common practice is to compare measurements of a series of different design linewidths on a given wafer spanning the range of smallest to largest dimensions expected to be encountered in the manufacturing line, and then minimize the average difference (offset) between the measurements of different tools. One shortcoming of this approach is that there is insufficient information to understand the root cause of an unacceptable average offset. Another approach attempts to have instruments to be matched produce data having a straight line with unity slope and zero intercept or average offset when comparing measurements of different design linewidths. This approach is an improvement in that the slope provides magnification error information but also suffers from the problem that insufficient diagnostic information is generated to identify root causes of unacceptable matching. In addition, both approaches fail to produce a comprehensive metric that combines all relevant matching information. Another shortcoming in current practices is the use of simplified artifacts for the matching measurements. Matching artifacts are often chosen because they are stable, reliably manufactured, and with little process-induced variation. Unfortunately, these very properties imply they are not leading edge technology examples nor do they display the full range of measurement challenges present in manufacturing.
In view of the foregoing, there is a need in the art for improved methods that address the problems of the related art.